A Triangulation Overview

Posted on by

A Segment-ology TIDBIT

Triangulation is a tool. It’s a process that can help us with our genealogy. It is not the only tool in our kit bag – there are many other tools that also utilize DNA, including InCommonWith Lists, Matching Segment Lists, Matrix displays, Shared Matches, Clustering, Circles, etc, etc. This blogpost is an overview of Triangulation.

With atDNA we have been using Triangulation to mean two different things:

Segment Triangulation of shared segments (a focus of this blog), and

Ancestry Triangulation (having at least 3 Matches in a Triangulated Group (TG) all match on the same ancestral line; sharing a Common Ancestor (CA) on that line.

In the atDNA community we often conflate these two concepts, and they are very much intertwined. I tend to think first of forming a TG and then looking at the genealogy to determine the side (maternal or paternal) and then finding various MRCAs. But some start with the genealogy and look for Triangulation to add evidence that a CA is correct. Both ways will work, they are intertwined in genetic genealogy, so in this overview I will also conflate them. Here are some overview points about Triangulation:

We look for at least 3 Matches, Much of our work as genealogists involves one-on-one – finding a Common Ancestor with a Match – that’s OK, but it’s not Triangulation.

We look at overlapping DNA segments. ICW and other tools don’t require overlapping segments – that’s OK, but they are not Triangulation.

We look for 3 “segment” legs. This means the 3 people (usually you and two Matches) that form a Triangulated Group are not closely related. But once a TG of 3 cousins is formed, other close relatives can be added to the TG. It’s the TG forming that needs 3 strong legs. So 3 siblings and their parent do not form a TG, but they can be in one.

The shared segments that form a TG must be IBD. From experience we’ve found that:

  • “all” shared segments over 15cM are IBD;
  • shared segments under about 7cM are false most of the time; and
  • the process of comparing overlapping shared segments in a TG will cull out many in the 7 to 15cM range which do not match – I consider these to be false segments.

Blaine Bettinger is working to define Triangulation – not to preclude the use of other tools – but to help us better understand Triangulation as a tool. I use Triangulation as a tool to primarily sort and group all of my IBD segments. I’ve formed about 400 separate TGs over my 45 chromosomes. New Matches always fall into one of these TGs (close Matches may span two or more TGs – it’s OK). This is segment Triangulation. With close relatives, I’ve been able to determine the side for these 400 TGs. This is a huge benefit because new Matches almost always Triangulate with other Matches already in a TG; and I then know which side our Common Ancestor must be on. This is an excellent use of the Triangulation tool.

Ancestry Triangulation does not preclude me from also using the information of Circles, or ICW lists, or ethnic makeup, of even genealogy records or Trees or discussions with Matches to determine CAs.

Within a TG we may find a CA with a Match. As we have pointed out many times: shared DNA plus a CA does NOT mean that the shared DNA came from that CA, or that the CA is somehow “proved” because there is also a shared segment – maybe, but also maybe not. But, by finding 3 Matches in a TG who all share the same CA (Ancestry Triangulation), we increase our confidence (not “prove”) that the CA is linked to the shared segment; and with more Ancestry Triangulation (and/or walking the ancestor back), we increase our confidence even more.

If Triangulation leads to a conclusion that a CA is not linked by the shared DNA, we can still be cousins on that CA, and we can still use ICW, Circles, etc. to pursue a genealogy goal. But we should not say that DNA supports that cousinship conclusion.

IMO, a TG has characteristics that help us in our genealogy goals. Triangulation is a strong tool that takes advantage of our shared DNA with Matches.

I applaud Blaine’s effort to try to define Triangulation and provide some standards for its use.

The above is adapted from my recent post to the Genetic Genealogy Tips and Techniques Facebook Group.

 

[22N] Segment-ology: A Triangulation Overview TIDBIT; by Jim Bartlett 20170728

25 thoughts on “A Triangulation Overview”

  1. Doug, WRT your 5 points – yes, the entire process is time consuming. It can be shortened somewhat by sorting your spreadsheet by cM and then running Triangulation on only the shared segements above, say, 15cM. Yes it takes time – IMO it’s worth it, because I wind up with a very useful tool – TGs forming a chromosome map. You can also save some time by only focusing on selected TGs formed with your closer, known, cousins.
    I agree that software tool are needed. Several have been produced – see DNAgedcom for instance, and GEDmatch, and more are in the works. The percent of chromosome mappers among the 8 million who have tested is very, very small, and in general the major companies won’t cater to us. They are always working on improvements – eg 23andMe’s ICW “Yes”; AncestryDNA’s Circles and Shared Matches. and FTDNA’s buckets.
    You are free to chase small segments if you want – how will you know they are IBD?
    Jim

    Like

    Reply ↓

    • A few more counter-points Jim:

      – on the testing companies not providing this tool: I would argue that the reason chromosome mappers are such a small segment (pardon the pun) of the GG community is precisely because the majority either have trouble understanding the intricacies, or simply do not have the time. Yes there are many who only want to spend minutes a week on their GG, but there’s a heck of a lot of genealogists who would spend hours a week… just not days a week. The hard truth is that MOST people who invest in a test lose interest precisely because the information presented is not sufficient to make good progress without a steep learning curve and a LOT of time. The needed algorithms are not rocket science (and I should know because I am one) and could be developed by a mid-level software engineer in a reasonable time period. The processing power needed on FTDNA’s end “might” be a challenge for them, in which case they could provide the tool to work from excel on one’s home computer… or they could limit the triangularization results to be reported only weekly or monthly until their processing power catches up. I find it curious why all the GG bloggers mostly take the PC line with the testing companies rather than calling them out for offering a test without the needed service. FTDNA seems to want to be like a medical laboratory that performs your blood test… but please go find a doctor for analysis. That model is understandable for the complexity discipline that is health care, but for GG, it is inexcusable. I simply do not understand why they continue on a business model that ultimately fails to generate meaningful results for their clients?

      – on chasing small segments and how will I know they will be IBD: first you know that one will not easily know whether a small segment is IBD or not… and you know the statistics on its likelihood already. In fact you don’t know that at 12cM segment that triangulates is not IBS… all you know is that the probabilities exponentially increase in one’s favour as the segments get larger – that’s no surprise. but I think I can apply the same logic that you are currently applying to larger segments: if 3 test-kit-triangularizaton occurs at a 20cM segment then we can be confident that the identified MRCA has a very high probability of being correct. Similarly, a 12cM quadrangularization offers similar certainty. therefore, if I have 9 test kits “nona-angularization” matching must provide similar certainly for a yet to be determined segment size – maybe 4cM. Furthermore, undertaking parental and grandparental phasing would be another very valuable tool to eliminating IBS… if only the tools were provided to allow one to construct missing genomes using several siblings, aunts or uncles. Such software is also possible to write, yet we are left to try to do this manually!

      And oh by the way, if FTDNA was providing such a multi-angularization tool, we would then be getting statistical analyses and probabilities of small segment matching. The reason we don’t have that statistical data is BECAUSE folks, like yourself, are being forced to do this manually and it is beyond practical means to extract such data. The truth is that the level of complexity needed to generate these lab results is well beyond what is needed to analyze the data contained in a single database. The testing companies don’t invest in the downstream service because their current mouse trap is sufficient for catching unsuspecting consumers.

      I wonder how many of the public would continue to test at FTDNA if their website highlighted in large text a caveat such as

      “POTENTIAL CUSTOMER PLEASE NOTE that the tools needed to extract the majority of the meaningful analysis contained in the data we provide is not offered herein. 3rd parties may offer some such tools, but are not granted access to our database to make such tools meaningful. You are nonetheless welcome to manually email thousands of other FTDNA customers and beg them, over many years, to also participate in downloading data from the FTDNA website and uploading it to another 3rd party website whose operators make no explicit claims of data integrity or safety. Regardless, please continue to encourage testing at our company as we all know in the GG community that it is only thru building a large database of testers will sufficient data exist to some day allow someone (maybe or maybe not us) to properly utilize this data. In the meantime, we do appreciate your contribution to FTDNA… crowd-sourcing genetic research can only continue with your participation.”

      Sorry for the ranting, but this is how I feel about all the testing companies. They hoodwink consumers to invest in tests knowing very well that the “goods” won’t and can’t be delivered by the tools they offer.

      Like

      Reply ↓

  2. Jim you make the process of triangularization sound simple, but in reality it seems quite a dog’s breakfast to me. On chromosome 1 for example, I have about 50 different matches that are active in the 40K to 90K address range. There is no clear start and stop points for most potential TG group (segments are nested inside each other, and overlap), and many matches are ICW others, but not uniformly. ie 1 & 2 are ICW 23 & 24. 1 & 23 are ICW 30. 23 & 30 are ICW 16, but 16 is not ICW 1 etc. And to complicate things, two matches that do not match on chromo 1, can in fact match on another chromosome. This may be easy when looking at large segments of the past few generations, but that is NOT where the frontier of GG is. I don’t need atDNA to tell me who my 2nd grt grandparents were… but rather to help me determine who their 2nd grt grandparents were! When looking at potential TGs from ancestors in the 1600s and 1700s, it becomes spaghetti very fast. In my case, my matches all have colonial ancestors in New England, Virginia, Carolinas etc.

    Question: do segments that all start or stop with the exact same address carry any special meaning wrt IBD? If I have 10 matches that all begin at say chr 16 43680, but each with varying segment length… say some with 23, some with 11, some with 8, some with 4, some with 2 cM – is there anything that can be implied from that? This group of 10 can form several potential “TGs”, with each TG having anywhere from 3 thru 10 members, depending on segment size.

    I think it is scandalous that FTDNA don’t provide a tool that allows comparison between two matches so as to confirm or refute triangularization! Combine that with 50% of matches not even uploading a tree, and all of this effort seems to futile.

    Like

    Reply ↓

    • From the DNA-EXPLAINED BLOG:

      And for the record, matches that begin and/or end in the same location are NOT more likely to be legitimate matches than those that start and end in different locations. Vendors use small buckets for matching, and if you fall into any part of the bucket, even if your match doesn’t entirely fill the bucket, the bucket is considered occupied. So what you’re seeing are the “fuzzy” bucket boundaries.

      Like

      Reply ↓

      • Doug, Every year so far, our list of Matches has about doubled. So the number of Matches in each TG has similarly doubled. Many TGs today, would have been called pile-ups before. Many of these pile-ups reflect distant Common Ancestors.

        Like

      • Thanks Jim for taking your time for all your feedback – it is appreciated. The concepts, as you describe, are clear to me, but the implementation seems not so straight forward.. and tedious is an under-statement – here’s my explanation of why I believe so (and please note these comments reflect frustration in the commercial testing companies and NOT in your great blog):

        1. a missing ICW between two individuals that are part of a larger group that appear to be forming a TG… but cannot yet be verified with a MRCA, makes the ICW process of verifying a TG very fuzzy. Maybe I should simply focus on small groups of 3 (me + 2 others) with pure ICWs and ignore that they are also overlapping with many other small groups who together collectively form only partial ICWs? Is this where I am making the mistake and thus complicating the effort?

        2. My “paper” tree is almost entirely filled out into the 1700s, but most of my matches do not have trees anywhere near that deep. In fact, about half choose not to post a tree at all. If I am to rely on individual emails to HUNDREDS… becoming THOUSANDS of atDNA matches at FTDNA, including requesting them to investigate match addresses in their chromosome browser (and assuming them to competently report back the correct conclusions), I will need this life and a few others to make adequate progress – noting that I am still fully employed!

        3. Concerning small segments, although I understand the statistics, it is here where TGs could add the value I seek. The forming of TGs from those distant ancestors is the very reason I have spent $thousands on testing! I have nonetheless found a compromise so that I can follow your advice: I am hiding all the small segments and trying to work only on the large segments first. However, thus far, few large segments reveal themselves to potential TGs (other than obvious 1st & 2nd cousin TGs)… and is thus why I was originally including the smaller segments… whose pile ups I sense may reflect the distant ancestors I seek.

        4. I have my myself and my mother also on AncestryDNA… but without chromosome data, I do not see why you raise the matter in this discussion? Ancestry.com is an ocean of erroneously copied trees (containing very little documentation) and no way to actually confirm a real TG. As you have often advocated, the fact that two people match and share a common “on paper” ancestor is of little value unless it can be triangulated… and Ancestry provides no tool to do that.

        5. GEDMATCH offers nice tools, but what percentage of my FTDNA, 23andMe and Ancestry matches are uploaded there? VERY LITTLE. And with all the whacky nicknames that people use, it is again futile to try to map GEMATCH individuals with names offered at the testing sites. Why people think it is cute and productive to post a variety of nicknames at different sites is beyond me.

        Anyway, I have no doubt that your efforts have helped form a clear chromosome map for yourself – and that it took likely 1000s of hours to accomplish. Surely this could be much more easily done with some simple software tools… (as demonstrated by GEDMATCH) yet ALL of the test providers refuse to offer any. That was acceptable several years ago, but today with the large numbers of testers and thus matches, it is quickly becoming an over-bearing exercise that is functionally futile.

        Honestly, how much work would it be to provide a direct TG tool at FTDNA? Hire a software engineer and it could be done within a few months. The bottom line is that FTDNA wants to be a genetic genealogy provider without providing any sharing information that would be deemed as personal. The two paradigms are not compatible. It is not unreasonable to ask users to accept that 3 persons in an ICW group have the ability to view each other’s chromosome matches.

        Like

      • Doug, If DNA-EXPLAINED quote is from my, I need to amend it. We should not be using “same location” start/end locations to analyze shared segments or make Triangulation decisions that way. Use the overlapping and matching criteria.
        FTDNA appears to use 100SNP “buckets”, and, as I understand it, if a Match’s segment doesn’t match the whole bucket, that bucket is not part of any shared segment. In other words, shared segments are made up of fully matching 100SNP buckets. This tends to shorten some FTDNA shared segments a little – but not very significantly. It does contribute to “fuzzy” TG boundaries.

        Jim

        Like

    • Doug, I think the Triangulation process with shared segments is fairly straightforward. It is also tedious and time consuming, particularly when you start. Now that I have TG which cover 98% of my DNA, new shared segments fall easily into place. Because of Colonial Ancestry, and random DNA, your ICW Matches can be all over the place. But only the ones who overlap each other can be Triangulated into TGs. Yes, an ancestor can pass down multiple segments of DNA to you – and each one will be part of a different TG – usually with different Matches. Looking at an entire genome, it can look messy. That’s the beauty of TGs – they represent only one ancestral line. So I usually work with one TG at a time.
      Your DNA is unique. The crossover points created by your ancestors, over many generations, are fixed points in your DNA. And each Match got their own segment from a Common Ancestor. What you’ll “see” in a Chromosome Browser, is the overlapping part of your and their DNA – the part of both your segments that is shared = the shared segment. Your Match may have a segment that starts before yours does, but all you’ll see is the overlap that can only start where your segment from the CA starts. So, yes, it’s expected that some shared segments will “stack up” at the start of your segment from the CA; and other shared segments may “stack up” where your segment from the CA ends. These points basically define your segment from the CA, and also the TG. In general, the longer the segment, the closer the cousinship – but clearly that is not a hard rule, and what you get often varies a lot.
      Caution – I would not include any 2cM or 4 cM segments in your analaysis – in fact since you are just starting out, I’d recommend form TGs only with segments over 10cM.
      Note: AncestryDNA has 10 times the Matches as FTDNA (I have 46,000 Matches at AncestryDNA!), and no segment data or browser. And they, too, do not allow comparison between Matches. In my blog I’ve posted a method for Triangulating at FTDNA.
      Jim

      Like

      Reply ↓

  3. I’m just learning about atDNA and triangulation. Can you tell me how you keep track of your TG’s? I can see the value of identifying TG’s and using them but I don’t really know how to go about keeping track of them. I’m just starting to use Genome Mate Pro and I could use the Group name but that doesn’t really tell me that it is a TG.
    Ideas?
    thanks
    Karen

    Like

    Reply ↓

    • Karen, I started by using a spreadsheet of Matches and then included the segment data. This was way before Genome Mate Pro, and I now have over 12,000 lines in my spreadsheet. But for many, Genome Mate Pro is a good way to handle all the data. They should have notes field that would allow you to enter a TG name. I use the NOTES box at Ancestry with entries like: 434M/7C1R: RUCKER/TINSLEY [02F36] 12.6cM This packs a lot of info into a shorthand: 434 is Ahnentafel for this RUCKER Ancestor; (435 is TINSLEY). I’ve added the M just to remind me it’s on my maternal side. 7C1R is 7th cousin once removed – each Match in a TG could be a different level of cousinship. [02F36] is my name for this TG. This Match is sharing 12.6cM with me. I can sort my spreadsheet by the TG name and see all the Matches, and shared segments, in one short list.

      Like

      Reply ↓

  4. A note that’s a bit tangential: Although I’m sure the readers of this blog understand this very well, one point that often seems to be missing these conversations is any recognition of the fact that triangulation is — in and of itself — a consideration of the possibility of DNA inheritance through a line other than the hypothesized common ancestor(s). In other words, we triangulate because we recognize the possibility of alternative shared ancestors. Although triangulation doesn’t eliminate the need to search conscientiously for all shared ancestors — it does, to some degree, address that issue directly.

    Like

    Reply ↓

  5. I certainly agree with you, Jim. On the issue of whether the term “triangulation” can be applied to autosomal comparisons involving fewer than three people… NO! Please let us reject this idea immediately. Although this conceptualization makes some sense with Y DNA comparisons (because the Y chromosome is not diploid), this conceptualization falls short with our diploid autosomal DNA. When working with autosomal DNA, we cannot rely on one-to-one comparisons to achieve what has been called “triangulation” with Y DNA.

    Like

    Reply ↓

  6. Great article. It helped clear up a few things for me. Regarding the 3 “segment” legs, are first cousins too close to make separate segment legs? You mentioned that parent and siblings would be considered one leg. It would seem that it might be counterproductive (or at least not of much use) were I to use first cousins on separate legs from that (assuming that the first cousins relate to everyone else on that leg and that they contain the segment I am working with, e.g., it’s the same MRCAs).

    Like

    Reply ↓

    • Elizabeth, you can use 1C. The idea is to form the TGs, and have them on both sides. I still “adjust” my TGs as more data (segments) come in. These are guidelines. And I’d add that it isn’t a requirement to only use known cousins – you can use any overlapping segments to form and fill up and fill out a TG. Most of your TGs will probably have more than just 3 segments.

      Liked by 2 people

      Reply ↓

      • Let me add that in using Triangulation for your own purposes, you had hedge on these guidelines – they are intended to significantly improve your odds. If you do hedge, you may have to backtrack and/or change more often as new data comes in. When I started with atDNA, I just focused on finding Common Ancestors – I’m a genealogist, what can I say. After about 2 years I had a list of about 100 Matches with Common Ancestors and put them into my first spreadsheet. I then started adding segment data, and learning about atDNA. It turns out about 1/4 of the Common Ancestors could not be right – they were on the wrong side; or they were on the same segment with a close cousin, but a different CA. A lot of work wasted because I didn’t use the DNA. And the other thing is if you want to state that an ancestor is linked to a DNA segment, you need to follow the generally accepted “guidelines” to promote a high confidence level that you are correct. I predict future “disputes” over ancestry (like we have today with paper genealogy), and the one who has followed the guidelines should win.

        Like

  7. So, I have a group of 40 matches all overlapping and sharing with each other averaging 12cM, range 10cM to 42cM, with only 4 matches above 15cM would you be discounting the other 36? Or, is the fact that so many share with each other of significance?

    Like

    Reply ↓

    • I’d keep them all. You never know – a 12cM Match may be a 4C with lots of research done on the CA; and the 20cM an adoptee, or no Tree, or not interested in genealogy, or 2C to someone else who paid for the kit and is not interested in helping any Match on other lines, or dead; or…. You never know where the gold is until you prospect for it.

      Like

      Reply ↓

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.